A step-down tranny is used. The secondary voltage is fed back to the wall voltage, out of phase, to subtract from, or
buck, the wall voltage. See:
http://sound.westhost.com/articles/buck-xfmr.htm and especially
http://geofex.com/ > Vintage AC adapter
This is much like using a variac, except the step-down is fixed, not variable.
Yes, heater voltage is the limiting factor. A nominal 6V heater can safely run on 5V (see Merlin's Valve Wizard site). Let's say 5
.3V for a margin of error; but is usually fed 6.3V: 5.3/6.3 = .84 So that's the largest factor that we can use to safely reduce supply voltage. Most PT specs assume 115VAC wall voltage X .84 = 97VAC. So that's the lowest voltage we want the bucking tranny (or a variac) to put-out. My wall voltage is actually about 123VAC. The 12.6V tranny should buck that down to about 110VAC. I'm actually getting 109VAC, well above 97VAC. So the math predicts the heaters will be safe. In fact, my DVM reads 6VAC on the heaters.
Also, I said
one thing I did was use the bucking tranny. The other thing I didn't mention is a 330Ω 25W inductive dropping resistor in the B+ rail. Under load, the bucking tranny and the power resistor drop about 50 plate volts each from the stock voltage to produce the overall 100V drop in the plate voltage.
I'm still toying with little bottle voltage downstream. This amp uses a voltage divider -- 22K series/150K shunt -- in the B+ rail to drop voltage to the PI & preamp. Removing the shunt resistor restores PI & preamp plate voltages to stock values.